sctp: Add Auto-ASCONF support (core).
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / sctp / protocol.c
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * Initialization/cleanup for SCTP protocol support.
12 *
13 * This SCTP implementation is free software;
14 * you can redistribute it and/or modify it under the terms of
15 * the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
17 * any later version.
18 *
19 * This SCTP implementation is distributed in the hope that it
20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
21 * ************************
22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
23 * See the GNU General Public License for more details.
24 *
25 * You should have received a copy of the GNU General Public License
26 * along with GNU CC; see the file COPYING. If not, write to
27 * the Free Software Foundation, 59 Temple Place - Suite 330,
28 * Boston, MA 02111-1307, USA.
29 *
30 * Please send any bug reports or fixes you make to the
31 * email address(es):
32 * lksctp developers <lksctp-developers@lists.sourceforge.net>
33 *
34 * Or submit a bug report through the following website:
35 * http://www.sf.net/projects/lksctp
36 *
37 * Written or modified by:
38 * La Monte H.P. Yarroll <piggy@acm.org>
39 * Karl Knutson <karl@athena.chicago.il.us>
40 * Jon Grimm <jgrimm@us.ibm.com>
41 * Sridhar Samudrala <sri@us.ibm.com>
42 * Daisy Chang <daisyc@us.ibm.com>
43 * Ardelle Fan <ardelle.fan@intel.com>
44 *
45 * Any bugs reported given to us we will try to fix... any fixes shared will
46 * be incorporated into the next SCTP release.
47 */
48
49 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
50
51 #include <linux/module.h>
52 #include <linux/init.h>
53 #include <linux/netdevice.h>
54 #include <linux/inetdevice.h>
55 #include <linux/seq_file.h>
56 #include <linux/bootmem.h>
57 #include <linux/highmem.h>
58 #include <linux/swap.h>
59 #include <linux/slab.h>
60 #include <net/net_namespace.h>
61 #include <net/protocol.h>
62 #include <net/ip.h>
63 #include <net/ipv6.h>
64 #include <net/route.h>
65 #include <net/sctp/sctp.h>
66 #include <net/addrconf.h>
67 #include <net/inet_common.h>
68 #include <net/inet_ecn.h>
69
70 /* Global data structures. */
71 struct sctp_globals sctp_globals __read_mostly;
72 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
73
74 #ifdef CONFIG_PROC_FS
75 struct proc_dir_entry *proc_net_sctp;
76 #endif
77
78 struct idr sctp_assocs_id;
79 DEFINE_SPINLOCK(sctp_assocs_id_lock);
80
81 /* This is the global socket data structure used for responding to
82 * the Out-of-the-blue (OOTB) packets. A control sock will be created
83 * for this socket at the initialization time.
84 */
85 static struct sock *sctp_ctl_sock;
86
87 static struct sctp_pf *sctp_pf_inet6_specific;
88 static struct sctp_pf *sctp_pf_inet_specific;
89 static struct sctp_af *sctp_af_v4_specific;
90 static struct sctp_af *sctp_af_v6_specific;
91
92 struct kmem_cache *sctp_chunk_cachep __read_mostly;
93 struct kmem_cache *sctp_bucket_cachep __read_mostly;
94
95 long sysctl_sctp_mem[3];
96 int sysctl_sctp_rmem[3];
97 int sysctl_sctp_wmem[3];
98
99 /* Return the address of the control sock. */
100 struct sock *sctp_get_ctl_sock(void)
101 {
102 return sctp_ctl_sock;
103 }
104
105 /* Set up the proc fs entry for the SCTP protocol. */
106 static __init int sctp_proc_init(void)
107 {
108 if (percpu_counter_init(&sctp_sockets_allocated, 0))
109 goto out_nomem;
110 #ifdef CONFIG_PROC_FS
111 if (!proc_net_sctp) {
112 proc_net_sctp = proc_mkdir("sctp", init_net.proc_net);
113 if (!proc_net_sctp)
114 goto out_free_percpu;
115 }
116
117 if (sctp_snmp_proc_init())
118 goto out_snmp_proc_init;
119 if (sctp_eps_proc_init())
120 goto out_eps_proc_init;
121 if (sctp_assocs_proc_init())
122 goto out_assocs_proc_init;
123 if (sctp_remaddr_proc_init())
124 goto out_remaddr_proc_init;
125
126 return 0;
127
128 out_remaddr_proc_init:
129 sctp_assocs_proc_exit();
130 out_assocs_proc_init:
131 sctp_eps_proc_exit();
132 out_eps_proc_init:
133 sctp_snmp_proc_exit();
134 out_snmp_proc_init:
135 if (proc_net_sctp) {
136 proc_net_sctp = NULL;
137 remove_proc_entry("sctp", init_net.proc_net);
138 }
139 out_free_percpu:
140 percpu_counter_destroy(&sctp_sockets_allocated);
141 #else
142 return 0;
143 #endif /* CONFIG_PROC_FS */
144
145 out_nomem:
146 return -ENOMEM;
147 }
148
149 /* Clean up the proc fs entry for the SCTP protocol.
150 * Note: Do not make this __exit as it is used in the init error
151 * path.
152 */
153 static void sctp_proc_exit(void)
154 {
155 #ifdef CONFIG_PROC_FS
156 sctp_snmp_proc_exit();
157 sctp_eps_proc_exit();
158 sctp_assocs_proc_exit();
159 sctp_remaddr_proc_exit();
160
161 if (proc_net_sctp) {
162 proc_net_sctp = NULL;
163 remove_proc_entry("sctp", init_net.proc_net);
164 }
165 #endif
166 percpu_counter_destroy(&sctp_sockets_allocated);
167 }
168
169 /* Private helper to extract ipv4 address and stash them in
170 * the protocol structure.
171 */
172 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
173 struct net_device *dev)
174 {
175 struct in_device *in_dev;
176 struct in_ifaddr *ifa;
177 struct sctp_sockaddr_entry *addr;
178
179 rcu_read_lock();
180 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
181 rcu_read_unlock();
182 return;
183 }
184
185 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
186 /* Add the address to the local list. */
187 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
188 if (addr) {
189 addr->a.v4.sin_family = AF_INET;
190 addr->a.v4.sin_port = 0;
191 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
192 addr->valid = 1;
193 INIT_LIST_HEAD(&addr->list);
194 list_add_tail(&addr->list, addrlist);
195 }
196 }
197
198 rcu_read_unlock();
199 }
200
201 /* Extract our IP addresses from the system and stash them in the
202 * protocol structure.
203 */
204 static void sctp_get_local_addr_list(void)
205 {
206 struct net_device *dev;
207 struct list_head *pos;
208 struct sctp_af *af;
209
210 rcu_read_lock();
211 for_each_netdev_rcu(&init_net, dev) {
212 __list_for_each(pos, &sctp_address_families) {
213 af = list_entry(pos, struct sctp_af, list);
214 af->copy_addrlist(&sctp_local_addr_list, dev);
215 }
216 }
217 rcu_read_unlock();
218 }
219
220 /* Free the existing local addresses. */
221 static void sctp_free_local_addr_list(void)
222 {
223 struct sctp_sockaddr_entry *addr;
224 struct list_head *pos, *temp;
225
226 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
227 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
228 list_del(pos);
229 kfree(addr);
230 }
231 }
232
233 /* Copy the local addresses which are valid for 'scope' into 'bp'. */
234 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
235 gfp_t gfp, int copy_flags)
236 {
237 struct sctp_sockaddr_entry *addr;
238 int error = 0;
239
240 rcu_read_lock();
241 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
242 if (!addr->valid)
243 continue;
244 if (sctp_in_scope(&addr->a, scope)) {
245 /* Now that the address is in scope, check to see if
246 * the address type is really supported by the local
247 * sock as well as the remote peer.
248 */
249 if ((((AF_INET == addr->a.sa.sa_family) &&
250 (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
251 (((AF_INET6 == addr->a.sa.sa_family) &&
252 (copy_flags & SCTP_ADDR6_ALLOWED) &&
253 (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
254 error = sctp_add_bind_addr(bp, &addr->a,
255 SCTP_ADDR_SRC, GFP_ATOMIC);
256 if (error)
257 goto end_copy;
258 }
259 }
260 }
261
262 end_copy:
263 rcu_read_unlock();
264 return error;
265 }
266
267 /* Initialize a sctp_addr from in incoming skb. */
268 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
269 int is_saddr)
270 {
271 void *from;
272 __be16 *port;
273 struct sctphdr *sh;
274
275 port = &addr->v4.sin_port;
276 addr->v4.sin_family = AF_INET;
277
278 sh = sctp_hdr(skb);
279 if (is_saddr) {
280 *port = sh->source;
281 from = &ip_hdr(skb)->saddr;
282 } else {
283 *port = sh->dest;
284 from = &ip_hdr(skb)->daddr;
285 }
286 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
287 }
288
289 /* Initialize an sctp_addr from a socket. */
290 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
291 {
292 addr->v4.sin_family = AF_INET;
293 addr->v4.sin_port = 0;
294 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
295 }
296
297 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
298 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
299 {
300 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
301 }
302
303 /* Initialize sk->sk_daddr from sctp_addr. */
304 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
305 {
306 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
307 }
308
309 /* Initialize a sctp_addr from an address parameter. */
310 static void sctp_v4_from_addr_param(union sctp_addr *addr,
311 union sctp_addr_param *param,
312 __be16 port, int iif)
313 {
314 addr->v4.sin_family = AF_INET;
315 addr->v4.sin_port = port;
316 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
317 }
318
319 /* Initialize an address parameter from a sctp_addr and return the length
320 * of the address parameter.
321 */
322 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
323 union sctp_addr_param *param)
324 {
325 int length = sizeof(sctp_ipv4addr_param_t);
326
327 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
328 param->v4.param_hdr.length = htons(length);
329 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
330
331 return length;
332 }
333
334 /* Initialize a sctp_addr from a dst_entry. */
335 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
336 __be16 port)
337 {
338 saddr->v4.sin_family = AF_INET;
339 saddr->v4.sin_port = port;
340 saddr->v4.sin_addr.s_addr = fl4->saddr;
341 }
342
343 /* Compare two addresses exactly. */
344 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
345 const union sctp_addr *addr2)
346 {
347 if (addr1->sa.sa_family != addr2->sa.sa_family)
348 return 0;
349 if (addr1->v4.sin_port != addr2->v4.sin_port)
350 return 0;
351 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
352 return 0;
353
354 return 1;
355 }
356
357 /* Initialize addr struct to INADDR_ANY. */
358 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
359 {
360 addr->v4.sin_family = AF_INET;
361 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
362 addr->v4.sin_port = port;
363 }
364
365 /* Is this a wildcard address? */
366 static int sctp_v4_is_any(const union sctp_addr *addr)
367 {
368 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
369 }
370
371 /* This function checks if the address is a valid address to be used for
372 * SCTP binding.
373 *
374 * Output:
375 * Return 0 - If the address is a non-unicast or an illegal address.
376 * Return 1 - If the address is a unicast.
377 */
378 static int sctp_v4_addr_valid(union sctp_addr *addr,
379 struct sctp_sock *sp,
380 const struct sk_buff *skb)
381 {
382 /* IPv4 addresses not allowed */
383 if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
384 return 0;
385
386 /* Is this a non-unicast address or a unusable SCTP address? */
387 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
388 return 0;
389
390 /* Is this a broadcast address? */
391 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
392 return 0;
393
394 return 1;
395 }
396
397 /* Should this be available for binding? */
398 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
399 {
400 int ret = inet_addr_type(&init_net, addr->v4.sin_addr.s_addr);
401
402
403 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
404 ret != RTN_LOCAL &&
405 !sp->inet.freebind &&
406 !sysctl_ip_nonlocal_bind)
407 return 0;
408
409 if (ipv6_only_sock(sctp_opt2sk(sp)))
410 return 0;
411
412 return 1;
413 }
414
415 /* Checking the loopback, private and other address scopes as defined in
416 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4
417 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
418 *
419 * Level 0 - unusable SCTP addresses
420 * Level 1 - loopback address
421 * Level 2 - link-local addresses
422 * Level 3 - private addresses.
423 * Level 4 - global addresses
424 * For INIT and INIT-ACK address list, let L be the level of
425 * of requested destination address, sender and receiver
426 * SHOULD include all of its addresses with level greater
427 * than or equal to L.
428 *
429 * IPv4 scoping can be controlled through sysctl option
430 * net.sctp.addr_scope_policy
431 */
432 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
433 {
434 sctp_scope_t retval;
435
436 /* Check for unusable SCTP addresses. */
437 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
438 retval = SCTP_SCOPE_UNUSABLE;
439 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
440 retval = SCTP_SCOPE_LOOPBACK;
441 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
442 retval = SCTP_SCOPE_LINK;
443 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
444 ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
445 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
446 retval = SCTP_SCOPE_PRIVATE;
447 } else {
448 retval = SCTP_SCOPE_GLOBAL;
449 }
450
451 return retval;
452 }
453
454 /* Returns a valid dst cache entry for the given source and destination ip
455 * addresses. If an association is passed, trys to get a dst entry with a
456 * source address that matches an address in the bind address list.
457 */
458 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
459 struct flowi *fl, struct sock *sk)
460 {
461 struct sctp_association *asoc = t->asoc;
462 struct rtable *rt;
463 struct flowi4 *fl4 = &fl->u.ip4;
464 struct sctp_bind_addr *bp;
465 struct sctp_sockaddr_entry *laddr;
466 struct dst_entry *dst = NULL;
467 union sctp_addr *daddr = &t->ipaddr;
468 union sctp_addr dst_saddr;
469
470 memset(fl4, 0x0, sizeof(struct flowi4));
471 fl4->daddr = daddr->v4.sin_addr.s_addr;
472 fl4->fl4_dport = daddr->v4.sin_port;
473 fl4->flowi4_proto = IPPROTO_SCTP;
474 if (asoc) {
475 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
476 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
477 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
478 }
479 if (saddr) {
480 fl4->saddr = saddr->v4.sin_addr.s_addr;
481 fl4->fl4_sport = saddr->v4.sin_port;
482 }
483
484 SCTP_DEBUG_PRINTK("%s: DST:%pI4, SRC:%pI4 - ",
485 __func__, &fl4->daddr, &fl4->saddr);
486
487 rt = ip_route_output_key(&init_net, fl4);
488 if (!IS_ERR(rt))
489 dst = &rt->dst;
490
491 /* If there is no association or if a source address is passed, no
492 * more validation is required.
493 */
494 if (!asoc || saddr)
495 goto out;
496
497 bp = &asoc->base.bind_addr;
498
499 if (dst) {
500 /* Walk through the bind address list and look for a bind
501 * address that matches the source address of the returned dst.
502 */
503 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
504 rcu_read_lock();
505 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
506 if (!laddr->valid || (laddr->state != SCTP_ADDR_SRC))
507 continue;
508 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
509 goto out_unlock;
510 }
511 rcu_read_unlock();
512
513 /* None of the bound addresses match the source address of the
514 * dst. So release it.
515 */
516 dst_release(dst);
517 dst = NULL;
518 }
519
520 /* Walk through the bind address list and try to get a dst that
521 * matches a bind address as the source address.
522 */
523 rcu_read_lock();
524 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
525 if (!laddr->valid)
526 continue;
527 if ((laddr->state == SCTP_ADDR_SRC) &&
528 (AF_INET == laddr->a.sa.sa_family)) {
529 fl4->saddr = laddr->a.v4.sin_addr.s_addr;
530 fl4->fl4_sport = laddr->a.v4.sin_port;
531 rt = ip_route_output_key(&init_net, fl4);
532 if (!IS_ERR(rt)) {
533 dst = &rt->dst;
534 goto out_unlock;
535 }
536 }
537 }
538
539 out_unlock:
540 rcu_read_unlock();
541 out:
542 t->dst = dst;
543 if (dst)
544 SCTP_DEBUG_PRINTK("rt_dst:%pI4, rt_src:%pI4\n",
545 &fl4->daddr, &fl4->saddr);
546 else
547 SCTP_DEBUG_PRINTK("NO ROUTE\n");
548 }
549
550 /* For v4, the source address is cached in the route entry(dst). So no need
551 * to cache it separately and hence this is an empty routine.
552 */
553 static void sctp_v4_get_saddr(struct sctp_sock *sk,
554 struct sctp_transport *t,
555 struct flowi *fl)
556 {
557 union sctp_addr *saddr = &t->saddr;
558 struct rtable *rt = (struct rtable *)t->dst;
559
560 if (rt) {
561 saddr->v4.sin_family = AF_INET;
562 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
563 }
564 }
565
566 /* What interface did this skb arrive on? */
567 static int sctp_v4_skb_iif(const struct sk_buff *skb)
568 {
569 return skb_rtable(skb)->rt_iif;
570 }
571
572 /* Was this packet marked by Explicit Congestion Notification? */
573 static int sctp_v4_is_ce(const struct sk_buff *skb)
574 {
575 return INET_ECN_is_ce(ip_hdr(skb)->tos);
576 }
577
578 /* Create and initialize a new sk for the socket returned by accept(). */
579 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
580 struct sctp_association *asoc)
581 {
582 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
583 sk->sk_prot);
584 struct inet_sock *newinet;
585
586 if (!newsk)
587 goto out;
588
589 sock_init_data(NULL, newsk);
590
591 sctp_copy_sock(newsk, sk, asoc);
592 sock_reset_flag(newsk, SOCK_ZAPPED);
593
594 newinet = inet_sk(newsk);
595
596 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
597
598 sk_refcnt_debug_inc(newsk);
599
600 if (newsk->sk_prot->init(newsk)) {
601 sk_common_release(newsk);
602 newsk = NULL;
603 }
604
605 out:
606 return newsk;
607 }
608
609 /* Map address, empty for v4 family */
610 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
611 {
612 /* Empty */
613 }
614
615 /* Dump the v4 addr to the seq file. */
616 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
617 {
618 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
619 }
620
621 static void sctp_v4_ecn_capable(struct sock *sk)
622 {
623 INET_ECN_xmit(sk);
624 }
625
626 void sctp_addr_wq_timeout_handler(unsigned long arg)
627 {
628 struct sctp_sockaddr_entry *addrw, *temp;
629 struct sctp_sock *sp;
630
631 spin_lock_bh(&sctp_addr_wq_lock);
632
633 list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
634 SCTP_DEBUG_PRINTK_IPADDR("sctp_addrwq_timo_handler: the first ent in wq %p is ",
635 " for cmd %d at entry %p\n", &sctp_addr_waitq, &addrw->a, addrw->state,
636 addrw);
637
638 /* Now we send an ASCONF for each association */
639 /* Note. we currently don't handle link local IPv6 addressees */
640 if (addrw->a.sa.sa_family == AF_INET6) {
641 struct in6_addr *in6;
642
643 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
644 IPV6_ADDR_LINKLOCAL)
645 goto free_next;
646
647 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
648 if (ipv6_chk_addr(&init_net, in6, NULL, 0) == 0 &&
649 addrw->state == SCTP_ADDR_NEW) {
650 unsigned long timeo_val;
651
652 SCTP_DEBUG_PRINTK("sctp_timo_handler: this is on DAD, trying %d sec later\n",
653 SCTP_ADDRESS_TICK_DELAY);
654 timeo_val = jiffies;
655 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
656 mod_timer(&sctp_addr_wq_timer, timeo_val);
657 break;
658 }
659 }
660
661 list_for_each_entry(sp, &sctp_auto_asconf_splist, auto_asconf_list) {
662 struct sock *sk;
663
664 sk = sctp_opt2sk(sp);
665 /* ignore bound-specific endpoints */
666 if (!sctp_is_ep_boundall(sk))
667 continue;
668 sctp_bh_lock_sock(sk);
669 if (sctp_asconf_mgmt(sp, addrw) < 0)
670 SCTP_DEBUG_PRINTK("sctp_addrwq_timo_handler: sctp_asconf_mgmt failed\n");
671 sctp_bh_unlock_sock(sk);
672 }
673 free_next:
674 list_del(&addrw->list);
675 kfree(addrw);
676 }
677 spin_unlock_bh(&sctp_addr_wq_lock);
678 }
679
680 static void sctp_free_addr_wq(void)
681 {
682 struct sctp_sockaddr_entry *addrw;
683 struct sctp_sockaddr_entry *temp;
684
685 spin_lock_bh(&sctp_addr_wq_lock);
686 del_timer(&sctp_addr_wq_timer);
687 list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
688 list_del(&addrw->list);
689 kfree(addrw);
690 }
691 spin_unlock_bh(&sctp_addr_wq_lock);
692 }
693
694 /* lookup the entry for the same address in the addr_waitq
695 * sctp_addr_wq MUST be locked
696 */
697 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct sctp_sockaddr_entry *addr)
698 {
699 struct sctp_sockaddr_entry *addrw;
700
701 list_for_each_entry(addrw, &sctp_addr_waitq, list) {
702 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
703 continue;
704 if (addrw->a.sa.sa_family == AF_INET) {
705 if (addrw->a.v4.sin_addr.s_addr ==
706 addr->a.v4.sin_addr.s_addr)
707 return addrw;
708 } else if (addrw->a.sa.sa_family == AF_INET6) {
709 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
710 &addr->a.v6.sin6_addr))
711 return addrw;
712 }
713 }
714 return NULL;
715 }
716
717 void sctp_addr_wq_mgmt(struct sctp_sockaddr_entry *addr, int cmd)
718 {
719 struct sctp_sockaddr_entry *addrw;
720 unsigned long timeo_val;
721
722 /* first, we check if an opposite message already exist in the queue.
723 * If we found such message, it is removed.
724 * This operation is a bit stupid, but the DHCP client attaches the
725 * new address after a couple of addition and deletion of that address
726 */
727
728 spin_lock_bh(&sctp_addr_wq_lock);
729 /* Offsets existing events in addr_wq */
730 addrw = sctp_addr_wq_lookup(addr);
731 if (addrw) {
732 if (addrw->state != cmd) {
733 SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt offsets existing entry for %d ",
734 " in wq %p\n", addrw->state, &addrw->a,
735 &sctp_addr_waitq);
736 list_del(&addrw->list);
737 kfree(addrw);
738 }
739 spin_unlock_bh(&sctp_addr_wq_lock);
740 return;
741 }
742
743 /* OK, we have to add the new address to the wait queue */
744 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
745 if (addrw == NULL) {
746 spin_unlock_bh(&sctp_addr_wq_lock);
747 return;
748 }
749 addrw->state = cmd;
750 list_add_tail(&addrw->list, &sctp_addr_waitq);
751 SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt add new entry for cmd:%d ",
752 " in wq %p\n", addrw->state, &addrw->a, &sctp_addr_waitq);
753
754 if (!timer_pending(&sctp_addr_wq_timer)) {
755 timeo_val = jiffies;
756 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
757 mod_timer(&sctp_addr_wq_timer, timeo_val);
758 }
759 spin_unlock_bh(&sctp_addr_wq_lock);
760 }
761
762 /* Event handler for inet address addition/deletion events.
763 * The sctp_local_addr_list needs to be protocted by a spin lock since
764 * multiple notifiers (say IPv4 and IPv6) may be running at the same
765 * time and thus corrupt the list.
766 * The reader side is protected with RCU.
767 */
768 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
769 void *ptr)
770 {
771 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
772 struct sctp_sockaddr_entry *addr = NULL;
773 struct sctp_sockaddr_entry *temp;
774 int found = 0;
775
776 if (!net_eq(dev_net(ifa->ifa_dev->dev), &init_net))
777 return NOTIFY_DONE;
778
779 switch (ev) {
780 case NETDEV_UP:
781 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
782 if (addr) {
783 addr->a.v4.sin_family = AF_INET;
784 addr->a.v4.sin_port = 0;
785 addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
786 addr->valid = 1;
787 spin_lock_bh(&sctp_local_addr_lock);
788 list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
789 sctp_addr_wq_mgmt(addr, SCTP_ADDR_NEW);
790 spin_unlock_bh(&sctp_local_addr_lock);
791 }
792 break;
793 case NETDEV_DOWN:
794 spin_lock_bh(&sctp_local_addr_lock);
795 list_for_each_entry_safe(addr, temp,
796 &sctp_local_addr_list, list) {
797 if (addr->a.sa.sa_family == AF_INET &&
798 addr->a.v4.sin_addr.s_addr ==
799 ifa->ifa_local) {
800 sctp_addr_wq_mgmt(addr, SCTP_ADDR_DEL);
801 found = 1;
802 addr->valid = 0;
803 list_del_rcu(&addr->list);
804 break;
805 }
806 }
807 spin_unlock_bh(&sctp_local_addr_lock);
808 if (found)
809 kfree_rcu(addr, rcu);
810 break;
811 }
812
813 return NOTIFY_DONE;
814 }
815
816 /*
817 * Initialize the control inode/socket with a control endpoint data
818 * structure. This endpoint is reserved exclusively for the OOTB processing.
819 */
820 static int sctp_ctl_sock_init(void)
821 {
822 int err;
823 sa_family_t family = PF_INET;
824
825 if (sctp_get_pf_specific(PF_INET6))
826 family = PF_INET6;
827
828 err = inet_ctl_sock_create(&sctp_ctl_sock, family,
829 SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
830
831 /* If IPv6 socket could not be created, try the IPv4 socket */
832 if (err < 0 && family == PF_INET6)
833 err = inet_ctl_sock_create(&sctp_ctl_sock, AF_INET,
834 SOCK_SEQPACKET, IPPROTO_SCTP,
835 &init_net);
836
837 if (err < 0) {
838 pr_err("Failed to create the SCTP control socket\n");
839 return err;
840 }
841 return 0;
842 }
843
844 /* Register address family specific functions. */
845 int sctp_register_af(struct sctp_af *af)
846 {
847 switch (af->sa_family) {
848 case AF_INET:
849 if (sctp_af_v4_specific)
850 return 0;
851 sctp_af_v4_specific = af;
852 break;
853 case AF_INET6:
854 if (sctp_af_v6_specific)
855 return 0;
856 sctp_af_v6_specific = af;
857 break;
858 default:
859 return 0;
860 }
861
862 INIT_LIST_HEAD(&af->list);
863 list_add_tail(&af->list, &sctp_address_families);
864 return 1;
865 }
866
867 /* Get the table of functions for manipulating a particular address
868 * family.
869 */
870 struct sctp_af *sctp_get_af_specific(sa_family_t family)
871 {
872 switch (family) {
873 case AF_INET:
874 return sctp_af_v4_specific;
875 case AF_INET6:
876 return sctp_af_v6_specific;
877 default:
878 return NULL;
879 }
880 }
881
882 /* Common code to initialize a AF_INET msg_name. */
883 static void sctp_inet_msgname(char *msgname, int *addr_len)
884 {
885 struct sockaddr_in *sin;
886
887 sin = (struct sockaddr_in *)msgname;
888 *addr_len = sizeof(struct sockaddr_in);
889 sin->sin_family = AF_INET;
890 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
891 }
892
893 /* Copy the primary address of the peer primary address as the msg_name. */
894 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
895 int *addr_len)
896 {
897 struct sockaddr_in *sin, *sinfrom;
898
899 if (msgname) {
900 struct sctp_association *asoc;
901
902 asoc = event->asoc;
903 sctp_inet_msgname(msgname, addr_len);
904 sin = (struct sockaddr_in *)msgname;
905 sinfrom = &asoc->peer.primary_addr.v4;
906 sin->sin_port = htons(asoc->peer.port);
907 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
908 }
909 }
910
911 /* Initialize and copy out a msgname from an inbound skb. */
912 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
913 {
914 if (msgname) {
915 struct sctphdr *sh = sctp_hdr(skb);
916 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
917
918 sctp_inet_msgname(msgname, len);
919 sin->sin_port = sh->source;
920 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
921 }
922 }
923
924 /* Do we support this AF? */
925 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
926 {
927 /* PF_INET only supports AF_INET addresses. */
928 return AF_INET == family;
929 }
930
931 /* Address matching with wildcards allowed. */
932 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
933 const union sctp_addr *addr2,
934 struct sctp_sock *opt)
935 {
936 /* PF_INET only supports AF_INET addresses. */
937 if (addr1->sa.sa_family != addr2->sa.sa_family)
938 return 0;
939 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
940 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
941 return 1;
942 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
943 return 1;
944
945 return 0;
946 }
947
948 /* Verify that provided sockaddr looks bindable. Common verification has
949 * already been taken care of.
950 */
951 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
952 {
953 return sctp_v4_available(addr, opt);
954 }
955
956 /* Verify that sockaddr looks sendable. Common verification has already
957 * been taken care of.
958 */
959 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
960 {
961 return 1;
962 }
963
964 /* Fill in Supported Address Type information for INIT and INIT-ACK
965 * chunks. Returns number of addresses supported.
966 */
967 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
968 __be16 *types)
969 {
970 types[0] = SCTP_PARAM_IPV4_ADDRESS;
971 return 1;
972 }
973
974 /* Wrapper routine that calls the ip transmit routine. */
975 static inline int sctp_v4_xmit(struct sk_buff *skb,
976 struct sctp_transport *transport)
977 {
978 struct inet_sock *inet = inet_sk(skb->sk);
979
980 SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n",
981 __func__, skb, skb->len,
982 &transport->fl.u.ip4.saddr,
983 &transport->fl.u.ip4.daddr);
984
985 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
986 IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
987
988 SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
989 return ip_queue_xmit(skb, &transport->fl);
990 }
991
992 static struct sctp_af sctp_af_inet;
993
994 static struct sctp_pf sctp_pf_inet = {
995 .event_msgname = sctp_inet_event_msgname,
996 .skb_msgname = sctp_inet_skb_msgname,
997 .af_supported = sctp_inet_af_supported,
998 .cmp_addr = sctp_inet_cmp_addr,
999 .bind_verify = sctp_inet_bind_verify,
1000 .send_verify = sctp_inet_send_verify,
1001 .supported_addrs = sctp_inet_supported_addrs,
1002 .create_accept_sk = sctp_v4_create_accept_sk,
1003 .addr_v4map = sctp_v4_addr_v4map,
1004 .af = &sctp_af_inet
1005 };
1006
1007 /* Notifier for inetaddr addition/deletion events. */
1008 static struct notifier_block sctp_inetaddr_notifier = {
1009 .notifier_call = sctp_inetaddr_event,
1010 };
1011
1012 /* Socket operations. */
1013 static const struct proto_ops inet_seqpacket_ops = {
1014 .family = PF_INET,
1015 .owner = THIS_MODULE,
1016 .release = inet_release, /* Needs to be wrapped... */
1017 .bind = inet_bind,
1018 .connect = inet_dgram_connect,
1019 .socketpair = sock_no_socketpair,
1020 .accept = inet_accept,
1021 .getname = inet_getname, /* Semantics are different. */
1022 .poll = sctp_poll,
1023 .ioctl = inet_ioctl,
1024 .listen = sctp_inet_listen,
1025 .shutdown = inet_shutdown, /* Looks harmless. */
1026 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1027 .getsockopt = sock_common_getsockopt,
1028 .sendmsg = inet_sendmsg,
1029 .recvmsg = sock_common_recvmsg,
1030 .mmap = sock_no_mmap,
1031 .sendpage = sock_no_sendpage,
1032 #ifdef CONFIG_COMPAT
1033 .compat_setsockopt = compat_sock_common_setsockopt,
1034 .compat_getsockopt = compat_sock_common_getsockopt,
1035 #endif
1036 };
1037
1038 /* Registration with AF_INET family. */
1039 static struct inet_protosw sctp_seqpacket_protosw = {
1040 .type = SOCK_SEQPACKET,
1041 .protocol = IPPROTO_SCTP,
1042 .prot = &sctp_prot,
1043 .ops = &inet_seqpacket_ops,
1044 .no_check = 0,
1045 .flags = SCTP_PROTOSW_FLAG
1046 };
1047 static struct inet_protosw sctp_stream_protosw = {
1048 .type = SOCK_STREAM,
1049 .protocol = IPPROTO_SCTP,
1050 .prot = &sctp_prot,
1051 .ops = &inet_seqpacket_ops,
1052 .no_check = 0,
1053 .flags = SCTP_PROTOSW_FLAG
1054 };
1055
1056 /* Register with IP layer. */
1057 static const struct net_protocol sctp_protocol = {
1058 .handler = sctp_rcv,
1059 .err_handler = sctp_v4_err,
1060 .no_policy = 1,
1061 };
1062
1063 /* IPv4 address related functions. */
1064 static struct sctp_af sctp_af_inet = {
1065 .sa_family = AF_INET,
1066 .sctp_xmit = sctp_v4_xmit,
1067 .setsockopt = ip_setsockopt,
1068 .getsockopt = ip_getsockopt,
1069 .get_dst = sctp_v4_get_dst,
1070 .get_saddr = sctp_v4_get_saddr,
1071 .copy_addrlist = sctp_v4_copy_addrlist,
1072 .from_skb = sctp_v4_from_skb,
1073 .from_sk = sctp_v4_from_sk,
1074 .to_sk_saddr = sctp_v4_to_sk_saddr,
1075 .to_sk_daddr = sctp_v4_to_sk_daddr,
1076 .from_addr_param = sctp_v4_from_addr_param,
1077 .to_addr_param = sctp_v4_to_addr_param,
1078 .cmp_addr = sctp_v4_cmp_addr,
1079 .addr_valid = sctp_v4_addr_valid,
1080 .inaddr_any = sctp_v4_inaddr_any,
1081 .is_any = sctp_v4_is_any,
1082 .available = sctp_v4_available,
1083 .scope = sctp_v4_scope,
1084 .skb_iif = sctp_v4_skb_iif,
1085 .is_ce = sctp_v4_is_ce,
1086 .seq_dump_addr = sctp_v4_seq_dump_addr,
1087 .ecn_capable = sctp_v4_ecn_capable,
1088 .net_header_len = sizeof(struct iphdr),
1089 .sockaddr_len = sizeof(struct sockaddr_in),
1090 #ifdef CONFIG_COMPAT
1091 .compat_setsockopt = compat_ip_setsockopt,
1092 .compat_getsockopt = compat_ip_getsockopt,
1093 #endif
1094 };
1095
1096 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1097
1098 switch (family) {
1099 case PF_INET:
1100 return sctp_pf_inet_specific;
1101 case PF_INET6:
1102 return sctp_pf_inet6_specific;
1103 default:
1104 return NULL;
1105 }
1106 }
1107
1108 /* Register the PF specific function table. */
1109 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1110 {
1111 switch (family) {
1112 case PF_INET:
1113 if (sctp_pf_inet_specific)
1114 return 0;
1115 sctp_pf_inet_specific = pf;
1116 break;
1117 case PF_INET6:
1118 if (sctp_pf_inet6_specific)
1119 return 0;
1120 sctp_pf_inet6_specific = pf;
1121 break;
1122 default:
1123 return 0;
1124 }
1125 return 1;
1126 }
1127
1128 static inline int init_sctp_mibs(void)
1129 {
1130 return snmp_mib_init((void __percpu **)sctp_statistics,
1131 sizeof(struct sctp_mib),
1132 __alignof__(struct sctp_mib));
1133 }
1134
1135 static inline void cleanup_sctp_mibs(void)
1136 {
1137 snmp_mib_free((void __percpu **)sctp_statistics);
1138 }
1139
1140 static void sctp_v4_pf_init(void)
1141 {
1142 /* Initialize the SCTP specific PF functions. */
1143 sctp_register_pf(&sctp_pf_inet, PF_INET);
1144 sctp_register_af(&sctp_af_inet);
1145 }
1146
1147 static void sctp_v4_pf_exit(void)
1148 {
1149 list_del(&sctp_af_inet.list);
1150 }
1151
1152 static int sctp_v4_protosw_init(void)
1153 {
1154 int rc;
1155
1156 rc = proto_register(&sctp_prot, 1);
1157 if (rc)
1158 return rc;
1159
1160 /* Register SCTP(UDP and TCP style) with socket layer. */
1161 inet_register_protosw(&sctp_seqpacket_protosw);
1162 inet_register_protosw(&sctp_stream_protosw);
1163
1164 return 0;
1165 }
1166
1167 static void sctp_v4_protosw_exit(void)
1168 {
1169 inet_unregister_protosw(&sctp_stream_protosw);
1170 inet_unregister_protosw(&sctp_seqpacket_protosw);
1171 proto_unregister(&sctp_prot);
1172 }
1173
1174 static int sctp_v4_add_protocol(void)
1175 {
1176 /* Register notifier for inet address additions/deletions. */
1177 register_inetaddr_notifier(&sctp_inetaddr_notifier);
1178
1179 /* Register SCTP with inet layer. */
1180 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1181 return -EAGAIN;
1182
1183 return 0;
1184 }
1185
1186 static void sctp_v4_del_protocol(void)
1187 {
1188 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1189 unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1190 }
1191
1192 /* Initialize the universe into something sensible. */
1193 SCTP_STATIC __init int sctp_init(void)
1194 {
1195 int i;
1196 int status = -EINVAL;
1197 unsigned long goal;
1198 unsigned long limit;
1199 unsigned long nr_pages;
1200 int max_share;
1201 int order;
1202
1203 /* SCTP_DEBUG sanity check. */
1204 if (!sctp_sanity_check())
1205 goto out;
1206
1207 /* Allocate bind_bucket and chunk caches. */
1208 status = -ENOBUFS;
1209 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1210 sizeof(struct sctp_bind_bucket),
1211 0, SLAB_HWCACHE_ALIGN,
1212 NULL);
1213 if (!sctp_bucket_cachep)
1214 goto out;
1215
1216 sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1217 sizeof(struct sctp_chunk),
1218 0, SLAB_HWCACHE_ALIGN,
1219 NULL);
1220 if (!sctp_chunk_cachep)
1221 goto err_chunk_cachep;
1222
1223 /* Allocate and initialise sctp mibs. */
1224 status = init_sctp_mibs();
1225 if (status)
1226 goto err_init_mibs;
1227
1228 /* Initialize proc fs directory. */
1229 status = sctp_proc_init();
1230 if (status)
1231 goto err_init_proc;
1232
1233 /* Initialize object count debugging. */
1234 sctp_dbg_objcnt_init();
1235
1236 /*
1237 * 14. Suggested SCTP Protocol Parameter Values
1238 */
1239 /* The following protocol parameters are RECOMMENDED: */
1240 /* RTO.Initial - 3 seconds */
1241 sctp_rto_initial = SCTP_RTO_INITIAL;
1242 /* RTO.Min - 1 second */
1243 sctp_rto_min = SCTP_RTO_MIN;
1244 /* RTO.Max - 60 seconds */
1245 sctp_rto_max = SCTP_RTO_MAX;
1246 /* RTO.Alpha - 1/8 */
1247 sctp_rto_alpha = SCTP_RTO_ALPHA;
1248 /* RTO.Beta - 1/4 */
1249 sctp_rto_beta = SCTP_RTO_BETA;
1250
1251 /* Valid.Cookie.Life - 60 seconds */
1252 sctp_valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE;
1253
1254 /* Whether Cookie Preservative is enabled(1) or not(0) */
1255 sctp_cookie_preserve_enable = 1;
1256
1257 /* Max.Burst - 4 */
1258 sctp_max_burst = SCTP_DEFAULT_MAX_BURST;
1259
1260 /* Association.Max.Retrans - 10 attempts
1261 * Path.Max.Retrans - 5 attempts (per destination address)
1262 * Max.Init.Retransmits - 8 attempts
1263 */
1264 sctp_max_retrans_association = 10;
1265 sctp_max_retrans_path = 5;
1266 sctp_max_retrans_init = 8;
1267
1268 /* Sendbuffer growth - do per-socket accounting */
1269 sctp_sndbuf_policy = 0;
1270
1271 /* Rcvbuffer growth - do per-socket accounting */
1272 sctp_rcvbuf_policy = 0;
1273
1274 /* HB.interval - 30 seconds */
1275 sctp_hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1276
1277 /* delayed SACK timeout */
1278 sctp_sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK;
1279
1280 /* Implementation specific variables. */
1281
1282 /* Initialize default stream count setup information. */
1283 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS;
1284 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS;
1285
1286 /* Initialize handle used for association ids. */
1287 idr_init(&sctp_assocs_id);
1288
1289 /* Set the pressure threshold to be a fraction of global memory that
1290 * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
1291 * memory, with a floor of 128 pages.
1292 * Note this initializes the data in sctpv6_prot too
1293 * Unabashedly stolen from tcp_init
1294 */
1295 nr_pages = totalram_pages - totalhigh_pages;
1296 limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1297 limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1298 limit = max(limit, 128UL);
1299 sysctl_sctp_mem[0] = limit / 4 * 3;
1300 sysctl_sctp_mem[1] = limit;
1301 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1302
1303 /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1304 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1305 max_share = min(4UL*1024*1024, limit);
1306
1307 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1308 sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
1309 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1310
1311 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1312 sysctl_sctp_wmem[1] = 16*1024;
1313 sysctl_sctp_wmem[2] = max(64*1024, max_share);
1314
1315 /* Size and allocate the association hash table.
1316 * The methodology is similar to that of the tcp hash tables.
1317 */
1318 if (totalram_pages >= (128 * 1024))
1319 goal = totalram_pages >> (22 - PAGE_SHIFT);
1320 else
1321 goal = totalram_pages >> (24 - PAGE_SHIFT);
1322
1323 for (order = 0; (1UL << order) < goal; order++)
1324 ;
1325
1326 do {
1327 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1328 sizeof(struct sctp_hashbucket);
1329 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1330 continue;
1331 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1332 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1333 } while (!sctp_assoc_hashtable && --order > 0);
1334 if (!sctp_assoc_hashtable) {
1335 pr_err("Failed association hash alloc\n");
1336 status = -ENOMEM;
1337 goto err_ahash_alloc;
1338 }
1339 for (i = 0; i < sctp_assoc_hashsize; i++) {
1340 rwlock_init(&sctp_assoc_hashtable[i].lock);
1341 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1342 }
1343
1344 /* Allocate and initialize the endpoint hash table. */
1345 sctp_ep_hashsize = 64;
1346 sctp_ep_hashtable = (struct sctp_hashbucket *)
1347 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1348 if (!sctp_ep_hashtable) {
1349 pr_err("Failed endpoint_hash alloc\n");
1350 status = -ENOMEM;
1351 goto err_ehash_alloc;
1352 }
1353 for (i = 0; i < sctp_ep_hashsize; i++) {
1354 rwlock_init(&sctp_ep_hashtable[i].lock);
1355 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1356 }
1357
1358 /* Allocate and initialize the SCTP port hash table. */
1359 do {
1360 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1361 sizeof(struct sctp_bind_hashbucket);
1362 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1363 continue;
1364 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1365 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1366 } while (!sctp_port_hashtable && --order > 0);
1367 if (!sctp_port_hashtable) {
1368 pr_err("Failed bind hash alloc\n");
1369 status = -ENOMEM;
1370 goto err_bhash_alloc;
1371 }
1372 for (i = 0; i < sctp_port_hashsize; i++) {
1373 spin_lock_init(&sctp_port_hashtable[i].lock);
1374 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1375 }
1376
1377 pr_info("Hash tables configured (established %d bind %d)\n",
1378 sctp_assoc_hashsize, sctp_port_hashsize);
1379
1380 /* Disable ADDIP by default. */
1381 sctp_addip_enable = 0;
1382 sctp_addip_noauth = 0;
1383 sctp_default_auto_asconf = 0;
1384
1385 /* Enable PR-SCTP by default. */
1386 sctp_prsctp_enable = 1;
1387
1388 /* Disable AUTH by default. */
1389 sctp_auth_enable = 0;
1390
1391 /* Set SCOPE policy to enabled */
1392 sctp_scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1393
1394 /* Set the default rwnd update threshold */
1395 sctp_rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1396
1397 sctp_sysctl_register();
1398
1399 INIT_LIST_HEAD(&sctp_address_families);
1400 sctp_v4_pf_init();
1401 sctp_v6_pf_init();
1402
1403 /* Initialize the local address list. */
1404 INIT_LIST_HEAD(&sctp_local_addr_list);
1405 spin_lock_init(&sctp_local_addr_lock);
1406 sctp_get_local_addr_list();
1407
1408 /* Initialize the address event list */
1409 INIT_LIST_HEAD(&sctp_addr_waitq);
1410 INIT_LIST_HEAD(&sctp_auto_asconf_splist);
1411 spin_lock_init(&sctp_addr_wq_lock);
1412 sctp_addr_wq_timer.expires = 0;
1413 setup_timer(&sctp_addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1414
1415 status = sctp_v4_protosw_init();
1416
1417 if (status)
1418 goto err_protosw_init;
1419
1420 status = sctp_v6_protosw_init();
1421 if (status)
1422 goto err_v6_protosw_init;
1423
1424 /* Initialize the control inode/socket for handling OOTB packets. */
1425 if ((status = sctp_ctl_sock_init())) {
1426 pr_err("Failed to initialize the SCTP control sock\n");
1427 goto err_ctl_sock_init;
1428 }
1429
1430 status = sctp_v4_add_protocol();
1431 if (status)
1432 goto err_add_protocol;
1433
1434 /* Register SCTP with inet6 layer. */
1435 status = sctp_v6_add_protocol();
1436 if (status)
1437 goto err_v6_add_protocol;
1438
1439 status = 0;
1440 out:
1441 return status;
1442 err_v6_add_protocol:
1443 sctp_v4_del_protocol();
1444 err_add_protocol:
1445 inet_ctl_sock_destroy(sctp_ctl_sock);
1446 err_ctl_sock_init:
1447 sctp_v6_protosw_exit();
1448 err_v6_protosw_init:
1449 sctp_v4_protosw_exit();
1450 err_protosw_init:
1451 sctp_free_local_addr_list();
1452 sctp_v4_pf_exit();
1453 sctp_v6_pf_exit();
1454 sctp_sysctl_unregister();
1455 free_pages((unsigned long)sctp_port_hashtable,
1456 get_order(sctp_port_hashsize *
1457 sizeof(struct sctp_bind_hashbucket)));
1458 err_bhash_alloc:
1459 kfree(sctp_ep_hashtable);
1460 err_ehash_alloc:
1461 free_pages((unsigned long)sctp_assoc_hashtable,
1462 get_order(sctp_assoc_hashsize *
1463 sizeof(struct sctp_hashbucket)));
1464 err_ahash_alloc:
1465 sctp_dbg_objcnt_exit();
1466 sctp_proc_exit();
1467 err_init_proc:
1468 cleanup_sctp_mibs();
1469 err_init_mibs:
1470 kmem_cache_destroy(sctp_chunk_cachep);
1471 err_chunk_cachep:
1472 kmem_cache_destroy(sctp_bucket_cachep);
1473 goto out;
1474 }
1475
1476 /* Exit handler for the SCTP protocol. */
1477 SCTP_STATIC __exit void sctp_exit(void)
1478 {
1479 /* BUG. This should probably do something useful like clean
1480 * up all the remaining associations and all that memory.
1481 */
1482
1483 /* Unregister with inet6/inet layers. */
1484 sctp_v6_del_protocol();
1485 sctp_v4_del_protocol();
1486 sctp_free_addr_wq();
1487
1488 /* Free the control endpoint. */
1489 inet_ctl_sock_destroy(sctp_ctl_sock);
1490
1491 /* Free protosw registrations */
1492 sctp_v6_protosw_exit();
1493 sctp_v4_protosw_exit();
1494
1495 /* Free the local address list. */
1496 sctp_free_local_addr_list();
1497
1498 /* Unregister with socket layer. */
1499 sctp_v6_pf_exit();
1500 sctp_v4_pf_exit();
1501
1502 sctp_sysctl_unregister();
1503
1504 free_pages((unsigned long)sctp_assoc_hashtable,
1505 get_order(sctp_assoc_hashsize *
1506 sizeof(struct sctp_hashbucket)));
1507 kfree(sctp_ep_hashtable);
1508 free_pages((unsigned long)sctp_port_hashtable,
1509 get_order(sctp_port_hashsize *
1510 sizeof(struct sctp_bind_hashbucket)));
1511
1512 sctp_dbg_objcnt_exit();
1513 sctp_proc_exit();
1514 cleanup_sctp_mibs();
1515
1516 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1517
1518 kmem_cache_destroy(sctp_chunk_cachep);
1519 kmem_cache_destroy(sctp_bucket_cachep);
1520 }
1521
1522 module_init(sctp_init);
1523 module_exit(sctp_exit);
1524
1525 /*
1526 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1527 */
1528 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1529 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1530 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1531 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1532 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1533 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1534 MODULE_LICENSE("GPL");